https://kubernetes.io/docs/tasks/configure-pod-container/assign-cpu-resource/ (or another page linked from there) should go into more detail on how CPU requests & limits actually work.

Specifically, on linux:

• Requests are translated to shares (cgroups v1) or weight (cgroups v2), and a brief overview of how the linux scheduler uses those
• Limits are translated to quota with a hardcoded period of 100ms, and a brief overview of how throttling works (in my experience, this is often misunderstood)

It would also be good to cover how windows handles these.

/sig node